Generative AI & Data Science Concepts

1. In data preprocessing, what is 'normalization'?

Removing duplicate rows from datasets

Scaling numerical features to a standard range

Converting categorical variables to numerical

Filling missing values with zeros

2. Which of the following is a key advantage of ensemble methods in machine learning?

They require less data than single models

Combining multiple models typically improves robustness and accuracy

They eliminate the need for hyperparameter tuning

They reduce computational requirements

3. What does 'backpropagation' do in neural network training?

Computes gradients and updates weights to minimize loss

Removes irrelevant neurons from the network

Validates the model on test data

Converts model outputs to probabilities

4. In neural networks, what is the purpose of an 'activation function'?

To add non-linearity to the network and enable learning complex patterns

To reduce the number of parameters in the model

To shuffle the training data

To convert categorical data to numerical form

5. What is the purpose of 'cross-validation' in machine learning?

To ensure the model works across different programming languages

To estimate model performance by training and testing on multiple data splits

To eliminate the need for a separate test set

To increase the size of the training dataset

6. What is the primary difference between supervised and unsupervised learning?

Supervised learning uses labeled data; unsupervised learning does not

Unsupervised learning is always more accurate

Supervised learning requires more computational power

Unsupervised learning only works with numerical data

7. In data science, what does 'feature engineering' involve?

Creating new features or transforming existing ones to improve model performance

Removing all categorical features

Standardizing column names

Deleting rows with missing values

8. What is 'hallucination' in the context of large language models?

When the model generates plausible but false or unsupported information

When the model refuses to answer questions

When the model perfectly recalls training data

When the model runs out of memory during inference

9. Which metric is most appropriate for evaluating a binary classification model?

Mean squared error only

Precision, recall, F1-score, or ROC-AUC

Perplexity

Mean absolute error only

10. What is the 'temperature' parameter in generative models used for?

Controlling the computational speed of inference

Adjusting the randomness or diversity of generated outputs

Measuring model accuracy on test data

Determining the number of training epochs

11. What is the primary advantage of transformer architecture over recurrent neural networks (RNNs) in processing sequential data?

Transformers require fewer parameters overall

Transformers use self-attention to process all positions in parallel

Transformers eliminate the need for embeddings

Transformers are exclusively used for image classification

12. What is the primary purpose of a 'validation set' in machine learning?

To train the final model

To evaluate model performance and tune hyperparameters without contaminating test results

To increase the size of training data

To replace the need for a test set

13. Which of the following best describes 'transfer learning'?

Copying weights from one model to create an unrelated model

Using knowledge from one task to improve performance on another task

Moving models between different servers

Converting TensorFlow models to PyTorch

14. What does 'fine-tuning' mean in the context of pre-trained language models?

Adjusting hyperparameters on the full dataset

Retraining a pre-trained model on task-specific data

Removing layers from the neural network

Increasing the vocabulary size

15. Which technique is commonly used to prevent a model from memorizing training data in generative AI?

Increasing batch size indefinitely

Dropout and regularization

Removing all validation data

Training for only one epoch

16. What is the primary function of an 'embedding' in natural language processing?

Converting words into high-dimensional numerical vectors

Removing stop words from text

Translating text between languages

Compressing text files

17. In machine learning, what does 'overfitting' indicate?

The model performs well on training data but poorly on unseen data

The model fails to learn any patterns

The dataset contains too many features

The model uses insufficient training data

18. What is the primary ethical concern associated with generative AI models regarding training data?

Training data requires too much computational power

Models may reproduce biases and copyrighted content from training data

Training data must always be labeled manually

Generative models cannot learn from any real data

19. Which of the following best describes the 'attention mechanism' in transformers?

A method to reduce model size by ignoring irrelevant data

A technique allowing the model to weigh the importance of different input tokens

A regularization technique to prevent overfitting

A process for compressing embeddings

20. In the context of large language models, what does the term 'prompt engineering' refer to?

Designing input text to elicit desired model outputs

Writing code to train neural networks

Optimizing server hardware for inference

Debugging model compilation errors

General Data Science Advanced Generative AI Concepts Quiz